Automated image segmentation-assisted flattening of atomic force microscopy images
Automated image segmentation-assisted flattening of atomic force microscopy images
复制标题
DOI：
10.3762/bjnano.9.91
复制
发表时间：
2018-03
影响因子：
3.1
通讯作者：
Yuliang Wang;T. Lu;Xiaolai Li;Huimin Wang
中科院分区：
文献类型：
--
作者：
Yuliang Wang;T. Lu;Xiaolai Li;Huimin Wang
原子力显微镜（AFM）图像通常呈现各种伪影。因此，在图像分析之前需要对图像进行平坦化处理。为了获得优化的平坦化结果，在图像平坦化过程中，一般采用矩形蒙版手工排除前景特征，耗时且不准确。在本研究中，提出了一种两步法自动实现优化的图像平坦化方案。在第一步中，通过精确的边界检测自动分割前景中的凸、凹特征；提取的前景特征作为排除蒙版。第二步，将背景中的数据点拟合为多项式曲线/曲面，然后从原始图像中减去这些多项式曲线/曲面，得到平坦的图像。此外，提出了基于滑动窗口的多项式拟合方法来处理具有复杂背景趋势的图像。介绍了两步图像平坦化方案的工作原理，研究了滑动窗口大小和多项式拟合方向对图像平坦化的影响。此外，利用该方法验证了图像平坦化对AFM图像形态表征和分割的作用。
Atomic force microscopy (AFM) images normally exhibit various artifacts. As a result, image flattening is required prior to image analysis. To obtain optimized flattening results, foreground features are generally manually excluded using rectangular masks in image flattening, which is time consuming and inaccurate. In this study, a two-step scheme was proposed to achieve optimized image flattening in an automated manner. In the first step, the convex and concave features in the foreground were automatically segmented with accurate boundary detection. The extracted foreground features were taken as exclusion masks. In the second step, data points in the background were fitted as polynomial curves/surfaces, which were then subtracted from raw images to get the flattened images. Moreover, sliding-window-based polynomial fitting was proposed to process images with complex background trends. The working principle of the two-step image flattening scheme were presented, followed by the investigation of the influence of a sliding-window size and polynomial fitting direction on the flattened images. Additionally, the role of image flattening on the morphological characterization and segmentation of AFM images were verified with the proposed method.